financial economics session 4: capm - curious quantcuriousquant.com/fecos/classes/session4.pdf ·...
TRANSCRIPT
Department of Economics
Financial Economics
Session 4: CAPM
Applicability, empirics and extensions
Postgraduate Class 2017
Economics Department Stellenbosch
Lecturer : Nico Katzke
News this week: 2014
• African Bank Investment Ltd (ABIL) had a rough old week…
• After the shock resignation of CEO Leon Kirkinis on
Wednesday (following an exceptionally poor trading statement
once again – a headline loss of over R6.4bn for the year to
September, requiring R8.5bn in new capital)– the share price
plummeted from R6.88 last Tuesday, to 30c by Friday…
shrinking the market value of ABIL from R10bn to R465mn in
one week!
• Although ABIL’s problems emerged last year with a report and
fine by the SA NCR.
2
News this week: 2014
• ABIL was yesterday placed under curatorship by the SARB.
• The SARB acted quickly and highly effectively in dealing with the
matter, as e.g. Kokkie Kooyman said:
• "I am impressed with the whole exercise and clarity of the Reserve Bank’s communication....
What they have done is bought time for the curator and managers to resolve outstanding
issues and eventually give shareholders a better basis on which to make a decision on
whether to put more capital on African Bank.
• The SARB split the Bank into a “Good bank” and a “Bad Bank” – buying the bad part
at a discount (paid R7bn for the R17bn part of the book – while the big four banks
and Investec and the PIC bought the “Good Part” of the book for R26bn.
• Trading of ABIL shares have been suspended on the JSE – and plans are being drawn
up to let it trade as the good bank part in time (to let investors opt in or out).
3
News this week: 2015
• Big News: China devalues its currency.
• This follows sustained strengthening of the dollar which has put strain on
the Chinese currency (Yuan) which largely tracks it.
• What does this imply?
• Could it signal structure weakness in the Chinese economy and that commodity
demand might be lower in the future?
• It might lead to the Fed postponing its rate hikes (as this would lead to a further
strengthening of the US$, and might lead to further relative Yuan depreciation –
which in turn could potentially hurt US trade competitiveness… Although this
VOX article suggests the US is in a better space today than it was in 2009, when
the Chinese were blamed for its competitive devaluation policies:
http://www.vox.com/2015/8/13/9149953/chinese-currency-devaluation-explained) 4
News this week: 2015
• Chinese devaluation stirs our market:
• The JSE had a rough ride this week following two days of consecutive
devaluations by Chinese authorities.
• Resources bounced back somewhat, but Naspers in particular took shots
due to its large Chinese market exposure…
• The two sides of the devaluation:
• China’s imports are hurt – impact on commodity prices and the rand.
• US’ relative export competitiveness might lead the Fed not to raise rates in
September following better economic environment, as this could lead to further
Chinese devaluations…
5
News this week
• Tencent rally brings valuation of Naspers into sharp perspective…
• Its one third holding of Chinese tech giant, Tencent, which saw excellent
results boost the share price tremendously this past year – is worth
roughly R1.8trn.
• Yet Naspers currently has a Market Cap of roughly R1.27trn.
• THUS: You are getting paid well to hold all the other companies in the
Naspers stable.
• What’s going on here (and has been for years)?!
• Is Naspers (at close to R3000 per share) undervalued, or is the tech
industry (particularly the fickle Chinese market) – experiencing a bubble?
6
ANNOUNCEMENT: Essay!
• The essay will be due last Friday of September.
• The following applies:
• Topic to be finalized and communicated to me the week after the term
break
• Needs to be finance focussed
• Maximum of 2500 words.
• Marks awarded based on quality of writing, depth of research, innovative
thinking / strength of reasoning, how well you addressed and covered your
topic
• Turnitin submission on or before the final date.
• The topic needs to be interesting and informative – the idea being that you teach
yourself something that falls outside the scope of this course, but which interests
you.
• See it as an opportunity to spend time tooling up on an area in finance that you are
interested in
• Read the FT, Bloomberg News, VoxEU, finance blogs, etc. to get some idea of what
interests you!
7
Plan for today
• CAPM’s empirical performance.
• Growth vs Value assets
• CAPM’s use in practice: How its values are
determined and by whom
• Extensions of CAPM
• Empirical performance.
• Anomalies in finance
8
CAPM’s major contribution
• The asset pricing equation that we derived last session, CAPM, provided investors
with an important insight – in that the expected excess return on holding a
security is directly proportional to its systematic (market) risk, as given
by its 𝛽, and is not related to its idiosyncratic (asset-specific) risk.
• This may sound completely un-intuitive… but consider this:
• Such idiosyncratic risk is not compensated for, as such risk is assumed to be
completely eliminated in an efficient market by combining the asset with other
imperfectly correlated assets in a diversified portfolio.
• Outperforming the market, therefore, requires investors to take on higher than unity
𝛽 assets in theory – thereby accepting higher risk in hope of getting higher return
𝑹𝒌 = 𝑹𝑹𝑭 + 𝜷𝒌 . (𝑹𝑴 − 𝑹𝑹𝑭)
9
How are the parameters used calculated in
practice?
• Several firms in SA offer services that provide key measures to their
clients with which to calculate the relevant risk-adjusted returns of
financial assets and portfolios.
• The one we will focus on today is that of the BNP Paribas report
(available on the course web page)
10
Portfolio’s Beta
• For many institutional investors and firms alike, calculating the weighted 𝛽 of a
portfolio’s shares is an important indication of the amount of risk the portfolio is
exposed to (for fund managers, investors and regulators alike).
• This also allows investors to assess the relative performance of a portfolio.
• The relative performance of a portfolio is an important concept that uninformed
investors tend to ignore.
• In the past (and even still today) many investors evaluated the performance of a
portfolio based on its returns only.
• This is, however, a dangerous exercise as risk-unadjusted performance measures
ignore the risks taken to achieve the return.
• A portfolio that has a higher risk exposure should, by definition, offer investors
higher returns – otherwise they are not compensated for the potential downside
to the investment strategies employed.11
Portfolio’s Beta
• The premium paid (i.t.o. fees) to asset managers that attempt to outperform
the market → are (in theory) justified by their ability to correctly predict the
movement of the market (or individual assets).
• Shifting the weighted Beta of a portfolio is vital in correctly “timing” the market
(move to high beta stocks in bull market / low beta stocks in bear markets…)
• Thus, in theory – in a closed universe where investors face only a
choice of risky equities and the RF asset – if a portfolio A has
weighted 𝜷 = 𝟏. 𝟓 and portfolio B has weighted 𝜷 = 𝟎. 𝟖, we expect
portfolio A to outperform portfolio B on aggregate, due to the
larger risk undertaken (very broadly speaking)
• If A earns 7% and B 6%, how do we assess which portfolio delivered the best
relative performance?12
Portfolio’s Beta
Comparing the risk-adjusted returns of different portfolios by:
1 Divide annual return by the portfolio Beta.
This is known as the Treynor Ratio:
T =𝑅𝑃−𝑅𝐹
𝛽𝑝𝑜𝑟𝑡𝑓𝑜𝑙𝑖𝑜
A higher Treynor ratio implies a portfolio having a higher risk-adjusted return.
The problem with this measure is that it implies that a portfolio is well diversified (so that 𝛽 =
good measure of the un-diversifiable risk the portfolio is exposed to)
2 We could also i.s.o. 𝛽 (which assumes a very well diversified portfolio) use the
Sharpe ratio
S =𝑅𝑃 − 𝑅𝐹𝜎𝑝𝑜𝑟𝑡𝑓𝑜𝑙𝑖𝑜13
Portfolio’s Beta
• Notice that the Sharpe ratio measures performance relative to the CML, while the
Treynor ratio measures performance relative to the SML.
• Although T is easier to compute (easy to find weighted 𝛽), S is a better comparable
figure (although it assumes the risk is accurately reflected in past variation…)
• In our closed universe, a weighted portfolio 𝛽 of 0.7 should by definition yield the
same as a portfolio consisting of 70% M and 30% RF assets.
• Another measure widely used to compare portfolios is Jensen’s Alpha:
• Here we compare two portfolios that have the same 𝜷, and determine their 𝛼’s.
• Remember from the previous session that 𝜶 was an indication of excess return
(above expected). Thus if true return = 𝑅𝑃, then:
Jensen’s Measure: 𝛼𝑃 = (𝑅𝑃 − 𝑅𝐶𝐴𝑃𝑀) = 𝑅𝑃 − {𝑅𝐹 − 𝛽𝑃 𝑅𝑀 − 𝑅𝐹 }14
Portfolio’s Beta
• Now suppose a portfolio manager earned a high 𝜶 in the last year.
• This implies that the return on his portfolio was higher than expected (or
above-average considering the risk-objectives)
• Is this in any way an indication of future positive and high 𝜶?
• Not even close.
• Perhaps it was luck. Or perhaps the manager hid some of the risk the
portfolio was exposed to (remember 2008 – many portfolio managers
scampered to stock up their funds with high earning “Super Safe” MBS
instruments… clearly belying their true portfolio risk and creating the
illusion of superior 𝛼′𝑠).
• Or perhaps (and I’m sure the manager will agree…), it was just down
to unique skill…15
What it then boils down to…
• Active investment strategies then boils down to chasing alphas.
• This chase can, however, yield an unfruitful exercise on aggregate → as the
costs erode the return on aggregate.
• The problem then is – in the short run your manager might have a high 𝛼, but
over the long term will probably return the market return – minus costs, it
would leave you in a negative sum game…
• See Bogel’s cynical, yet insightful paper on investments, as well as Nassim
Taleb’s interesting insight on the difference between noise and
information - posted on the site… is chasing alpha based on noisy
patterns a wise thing? 16
How are these measures computed in
practice?
• As mentioned, we will now look at how industry calculates the parameters
that we can use to evaluate assets and portfolios.
• First, the monthly (or whichever you have available) returns are calculated
for each of the relevant shares on the JSE, while the corresponding returns
of the JSE/FTSE ALSi (All Share Index) are also calculated.
• They then regress the two series as follows (For Anglo Am & JSE ALSi):
17
Source:
Cadiz:
Equity Risk Service
on the JSE
How are these measures computed in
practice?
• The individual shares and separate indexes (including, among many, the JSE Top 40-,
Financial-, Venture Capital-, AltX-, Listed Property-, Value- & Growth Indexes and also
all the shares listed) are then regressed relative to the Benchmark
• The Benchmarks in the report are: the JSE/FTSE ALSi, Financial and Industrial Index, Resource
top 20 Index & the JSE Top 40 index.
• Several key indicators are then provided for stock / indexes relative to the Benchmark
used:
1. 𝛃 → the regression line fitted using the regression techniques: This is (as hammered home
by now) an indication of the sensitivity of a share / index’s price change relative to the
benchmark (market).
2. Annualised 𝜶 → Avg return if the Benchmark had a zero change (thus the intercept of the
regression line): proxying excess returns for holding this share…
18
How are these measures computed in
practice?
3. SE(𝛃) → This is an indication off the reliability of the estimate of β. This is used to construct a
C.I. that we assume with 95% assurance contains the true β. By definition, a low SE indicates an
accurate estimate of β.
4. Annualized Total risk = 𝜎Returns : This measures the share / index’s total risk in % (but
remember our contention about risk measured as % deviation of returns…). To make the
monthly adjustment – divide the % by 12.
5. Annualized Unique risk = Fluctuations due to events unique to the company (strikes, bad
management, law suits…)
6. 𝐑𝟐 → Proportion of total risk accounted for by the market (systematic) risk.
A high 𝑹𝟐 implies much of the variation in Assets returns are caused by the Market’s
variation. Low 𝑹𝟐 → large proportion of movements due to unique factors
19
A few such examples of
Indexes relative to JSE ALSi (Dec 2014)
20
Expected total
monthly
variation:
72.4%/12 =6%
Index 𝜷 SE(𝜷) Ann Total
Risk (𝝈%)
Ann
Unique
Risk (%)
𝑹𝟐
JSE ALSi 1.00 0.00 18.7 0 100%
Top 40 1.06 0.01 18.7 1.7 99%
Resource
top 20
1.3 0.08 26.7 11 83%
Industrial
top 25
0.78 0.07 17.3 9.2 71%
Financial
top 15
0.74 0.11 21.1 15.8 43%
Venture
Capital
Index
1.5 0.49 72.4 66.2 15%
Alt X 0.61 0.17 26.1 23.3 19%
Value
Index
0.94 0.04 18.3 5.1 93%
Growth
Index
1.07 0.04 20.8 5.4 92%
A few such examples of
Shares relative to JSE ALSi (Dec 2014)
21
Share 𝜷 SE(𝜷) Ann Total
Risk (𝝈%)
Ann
Unique
Risk (%)
𝑹𝟐
ABSA 0.56 0.16 24.5 22 18%
FNB 0.78 0.18 28.9 24.7 25%
Nedbank 0.65 0.16 25.9 22.7 22%
Standard
Bank
0.74 0.17 28 24.1 25%
Capitec 0.63 0.2 30.2 27.6 15%
Old
Mutual
1.18 0.22 33.6 25.1 43%
Sanlam 0.81 0.14 25.1 19.8 37%
Anglo Am 1.67 0.16 37.8 21.2 68%
Shoprite 0.38 0.17 24.1 22.8 8%
MTN 0.84 0.16 27.1 22 34%
Vodacom 0.3 0.22 17.9 17.1 6%
Sasol 1.05 0.11 24.6 14.8 63%
Considerations when using such quantitative
indicators
• For shares that are thinly traded, the estimate of 𝛽 will be less accurate – the
Cadiz report controls for this.
• Time horizon → the report uses 5 years of monthly data (60 data points),
which might be subject to small sample biases – but the trade-off is that a
monthly frequency is subject to less noise and errors than higher frequency
(weekly / daily).
• Stability of Betas → as mentioned before, Betas can and do change over time.
This can be problematic for investors hoping to “cash in” on a high beta asset in
a bull market, only to find a dampened upside and an intensified downside made
the Beta look large…
• (for more details on the practitioners guide to calculating this measure… See:
EstBeta.pdf on webpage)
22
Considerations when using such quantitative
indicators
• As these indicators are based on past co-movements of the shares and the market, several factors
may influence its accuracy in determining future co-movements:
• Accuracy of past data – can be questioned as the past five years include a heavily distorting
global financial crisis.
• Intensified interconnectedness of certain shares in your portfolio at times, that negate the
diversification benefits and causes the portfolio to vary less than in accordance to the market.
• Unexpected events / unique factors of a business might become more pronounced in the
future – thus investors should be aware of this and keep it in consideration. This emphasizes
the need for qualitative evaluations and forecasted quantitative evaluations too.
• Some businesses listed on the JSE might be more influenced by external factors /
macroeconomic factors that do not necessarily influence the market. This can be seen by
viewing the low 𝑅2, which may not necessarily be a good indication of this.
23
Market Beta = Naspers?!
• Today the Beta measure has often been criticized
as implying the comovement largely with Naspers
(and a few other stocks).
• Which, indirectly, implies the comovement with the
Chinese market via tencent.
• A couple of years ago Beta was tilted towards the
resource sector (and thus showed high
comovement with commodity prices).
24
Examples of how we can use the indicators.
• Suppose you were provided with the tables above as given.
• Calculate the Beta and Risk of a portfolio that consists of the following asset
distribution (the nice thing about finding the 𝛽𝑃𝑜𝑟𝑡𝑓𝑜𝑙𝑖𝑜 is that it can simply be
added with the respective weightings):
25
SHARE % of portfolio
ABSA 30
ANGLO AMERICAN 20
MTN 10
JSE ALSi 30
R157 Bond 10
Portfolio Beta : σ(Weight) . (𝛽𝑆ℎ𝑎𝑟𝑒𝑠)
𝛽𝑃 = 0.3 0.56 + 0.2 1.67 + 0.1 0.84 + 0.3 1 + 0.1 0 = 0.886
Examples of how we can use the indicators.
Calculate the total Portfolio Risk:
Total Risk = Systematic (market) Risk + Non-Systematic (unique) Risk
𝑇𝑜𝑡𝑎𝑙 𝑅𝑖𝑠𝑘2 = 𝑀𝑎𝑟𝑘𝑒𝑡 𝑅𝑖𝑠𝑘2 + 𝑈𝑛𝑖𝑞𝑢𝑒 𝑅𝑖𝑠𝑘2
𝑇𝑜𝑡𝑎𝑙 𝑅𝑖𝑠𝑘2 = 𝑅2 𝑇𝑜𝑡𝑎𝑙 𝑅𝑖𝑠𝑘 2 + (1 − 𝑅2) 𝑇𝑜𝑡𝑎𝑙 𝑅𝑖𝑠𝑘 2
∴ Market risk of this proposed portfolio:
Market risk = 𝛽𝑃 . 𝑇𝑜𝑡𝑎𝑙 𝑟𝑖𝑠𝑘 𝑜𝑓 𝑡ℎ𝑒 𝑀𝑎𝑟𝑘𝑒𝑡 𝑖𝑛𝑑𝑒𝑥 = 𝛽𝑃 . (𝜎𝐽𝑆𝐸_𝐴𝐿𝑆𝑖)
= 0.886 18.7 = 𝟏𝟔. 𝟓𝟔%
Unique Risk of the portfolio: (bit more complicated)
Unique Risk = σ(𝑊𝑒𝑖𝑔ℎ𝑡). ( 1 − 𝑅2 ∗ 𝐴𝑛𝑛 𝑇𝑜𝑡𝑎𝑙 𝑅𝑖𝑠𝑘 2])
= 0.3 ∗ [ 1 − 0.18 . 24.5 ]2 + 0.2 * [ 1 − 0.68 . 37.8 ]2
+0.1* [(1 − 0.34) 27.1 ]2 = 𝟏𝟑. 𝟐%
26
Calculating the total Portfolio Risk:
27
Total Risk = Systematic (market) Risk + Non-Systematic (unique) Risk
𝑇𝑜𝑡𝑎𝑙 𝑅𝑖𝑠𝑘2 = 𝑀𝑎𝑟𝑘𝑒𝑡 𝑅𝑖𝑠𝑘2 + 𝑈𝑛𝑖𝑞𝑢𝑒 𝑅𝑖𝑠𝑘2
= 16.562 + 13.22 = 448.4
Thus the Total Risk of the portfolio that we proposed would be:
448.4 = 𝟐𝟏. 𝟏𝟕% 𝑷.𝑨
And the portfolio’s 𝑅2 =𝑀𝑎𝑟𝑘𝑒𝑡 𝑟𝑖𝑠𝑘
𝑇𝑜𝑡𝑎𝑙 𝑝𝑜𝑟𝑡𝑓𝑜𝑙𝑖𝑜 𝑅𝑖𝑠𝑘=
16.562
448.4= 0.61
& ∴ 1 − 𝑅2 = 0.39
So that the movement in the JSE ALSi explains approx. 𝟔𝟏% of the movement in our portfolio
– implying great scope left to diversify portfolio further & make the 1 − 𝑅2 statistic less (and
thus be less exposed to unique risks!!)
Abnormal Portfolio return
• Suppose you are given the following returns information at the end of the year.
• Calculate the portfolio’s abnormal return:
Remember: 𝛽𝑃 = 0.886
This implies the benchmark risk-adjusted return (i.e. of a portfolio having 88.6%
invested in JSE ALSi and 11.4% invested in R157 Bond) is:
0.886 8% + 1 − 0.886 5.5% = 𝟕. 𝟕𝟏𝟓%
28
SHARE Return
ABSA 15 %
ANGLO AMERICAN -3 %
MTN 10 %
JSE ALSi (𝑅𝑀) 8 %
R157 Bond (𝑅𝑅𝐹) 5.5 %
Abnormal Portfolio return
• Our portfolio return would then be:
0.3 15% + 0.2 −3% + 0.1 10% + 0.3 8% + 0.1 5.5% = 𝟕. 𝟖𝟓%
The Abnormal Return of our portfolio is thus:
7.85% − 7.715% = 0.135%
29
SHARE Portfolio Weight Return
ABSA 30% 15 %
ANGLO
AMERICAN
20% -3 %
MTN 10 % 10 %
JSE ALSi 30% 8 %
R157 Bond 10% 5.5 %
Abnormal Portfolio return
• Before applauding the portfolio manager too much, perhaps this
abnormal return of 𝟎. 𝟏𝟑𝟓% needs to be put into perspective…
• Although the Market risk of our Portfolio and the Benchmark portfolio
(on the SML in the CAPM) is identical, our portfolio has more unique
risk (𝟏𝟑. 𝟐% compared to the 𝟎% of the benchmark portfolio having
88.6% in M and 11.4% in RF).
• This implies that investors are merely compensated (with 0.135%) for
holding more unique risk in a less diversified portfolio…
• This follows as there is no compensation for holding unique risk
→ the core of the CAPM insights!!
30
Abnormal Portfolio return
• We can now calculate the two measures mentioned earlier:
T =𝑅𝑃 − 𝑅𝐹𝛽𝑝𝑜𝑟𝑡𝑓𝑜𝑙𝑖𝑜
=7.85 − 5.5
0.886= 2.65
S =𝑅𝑃 − 𝑅𝐹𝜎𝑝𝑜𝑟𝑡𝑓𝑜𝑙𝑖𝑜
=7.85 − 5.5
21.17= 0.11
• As mentioned earlier, however, due to the portfolio not being
highly diversified – it is better to use the Sharpe ratio in this
case (as the 𝛽 used in T assumes a sufficiently diversified portfolio).
31
CAPM’s relevance in SA
• CAPM, most likely due to its simplicity and applicability – is still
very much a useful starting point for any quantitative analysis of a
financial asset’s value, and forms the basis for more advanced
techniques.
• In fact, according to Correia and Cramer (2008: 41) the vast
majority of companies in SA use CAPM (or some basic derivative
thereof) to determine the fair cost of equity.
• The extensions thereof, such as APT and the dividend discount model,
are almost never used in the valuation of equity in practice
• The authors indicate a great movement toward explicitly valuing
equity i.t.o. CAPM (both in SA and abroad) over the last 15 years.
33
CAPM’s relevance in SA: RF
• What do firms use as the Risk Free rate?
• Correia and Cramer found the choice of bond yields to use in acting as a
proxy for the RF is quite varied. Although McKinsey & Co. recommend the
use of longer term bond yields that are less volatile, firms used a great
variation of bond yields (such as the R157 (maturing 2015), R201 (2014),
R186 (2026)*,,…) as a proxy for the RF. In fact, SA firms have shown to be
more inclined to use the shorter term yield (R157).
• Clearly there are no clear consensus of which bond yield to use as RF
proxy (which of course changes the pricing of equity between firms…
necessarily a bad thing for market liquidity?)
• *These are names of bonds with different maturities*
• We will be covering it in more detail in a later session, but for now you can check
out table 11.4 in the textbook for a more thorough indication.
34
CAPM’s relevance in SA: Equity premium
• The equity market premium of the CAPM: (𝑹𝑴 − 𝑹𝑹𝑭), reflects the
additional return that investors require for investing in the market
portfolio (or for holding equity in general).
• As with the proxy for RF, there is no clear consensus as to the calculation
of the equity premium for SA.
• Firms in SA have shown to use an equity premium of roughly 5.35%, while
the median is 5%.
• This is more or less consistent with the literature, with Kruger (2005)
estimating it at 5 – 5.5% and Dimson, et al (2003) at 5.2% historically.
35
CAPM’s relevance in SA: 𝛽
• Calculating accurate Betas of firms in SA seems an onerous task, considering all the
pitfalls and caveats..
• Lack of liquidity (of smaller firms whose assets do not trade as often as larger firms), arbitrary
time horizons and random occurrences that may not repeat in the future all cause
inaccuracies of the calculations of 𝛽.
• As seen earlier, firms can use the services of Cadiz / Bloomberg / McGregor / various other
sources – to calculate SA firm Betas .
• The most popular are Cadiz (25%), Bloomberg (25%) and McGregor (20%), while less
than 5% make in-house calculations].
• Correia & Cramer (2007)’s survey does, however, suggest that 44% of companies do
make some adjustments to their supplied Betas.
• The market portfolio (M) that is most widely used is the JSE ALSi (77%), but some firms see
this index as over-weighted toward large institutions (that assert strong individual influence on
the market outcome) and rather use the Financial and Industrial Index (FINDI) (roughly 23%) 36
CAPM’s relevance in SA
• Correia & Cramer (2008)’s survey therefore indicate that
SA firms “mostly adhere to practices that are consistent
with finance theory” – which reflects a highly developed
corporate and financial sector comparable with the US.
• They also suggest widespread use of key financial ratios in
determining the value of firms. This is important in order to
provide further quantitative assessments of the fundamental
valuations of firms.
• These ratios, and other important aspects, will be discussed
in two sessions’ time.37
Backward looking syndrome
• As became clear again following the recent financial turmoil, assuming
normality of returns in stock markets and implicitly assuming that what
happened in the past is a good indication of what will happen in the
future are two of the most pertinent problems with modern financial
modelling.
• The CAPM (and many subsequent models of which it’s theory is at the
core), does not describe or predict the source of risk (i.e. it does not
specify what causes the risk premiums), but merely predicts its
correlation with the market return, and then provides from it a forecast
basis for future risk premiums.39
Introducing the second APM:
Arbitrage Pricing Theory (APT)
• The APT approach seeks to remedy this problem by specifying various
independent macro-economic variables which predict certain sources of
risk (and hence deducing the fair price from it).
• APT – Ross (1976):
• Uses linear Factor Analysis to identify common factors in the
risk of security returns
• These factors are used to explain the unsystematic (or
idiosyncratic) risk of securities
• Returns do not have to be assumed jointly normally distributed
as this is achieved through the linearity of the factor analysis
(addresses some shortcoming of the CAPM)
40
APT – underlying assumptions
• It assumes that returns can be described in terms of a linear factor model.
• In this model, the return on any security, k, is dependent on many factors, not just the
market ( having in total 𝐺 independent factors that influence the random return of
asset 𝑘) :
41
(g)factor of (risk)y volatilit theto
return s)'(security theofy sensitivit the ][ :With
,~~...
~~
,
,11,
k
kffRR
gk
kGGkkMk
෨𝑅𝑘 → Expected return of (k)
𝑅𝑀 → Expected market return෩𝑓1 → deviation of factor 1 from its expected (mean)value
→ ෩𝑓1 > 0 is thus a shock that is not priced in (𝐸(෩𝑓1) = 0)
ǁ𝜀 → Noise, which we assume can be diversified away!
APT further….
• To simplify we use (the concept of) factor-specific ‘mimicking portfolios’
(i.e. create portfolios which mimic the key factors)
• Hence a portfolio is created which is perfectly correlated with the specific factor, but
has zero correlation with all the other factors in the model.
• This is because the model requires independent factors as causal variables
• The mimicking portfolio then has a sensitivity, 𝛽𝑔, of (close to) 1 to the specific factor
[𝑔], and (close to) 0 to the other factors in the model
• We view the deviations as the returns to these portfolios…
• NB: this is a functional, not a causal relationship
• The factors (or mimicking portfolios) are only correlated with the measured asset 𝑘’s
returns – they don’t necessarily cause the returns
• They are usually macro-type factors42
][ ggg FFf ~~
Example of APT: Macro-Factors
• An example of this could be: if asset (𝑘) is an Anglo American (gold
mining company) share, while factors influencing it could be:
• 𝑔 = (𝐺𝑜𝑙𝑑 𝑃𝑟𝑖𝑐𝑒, 𝑅/$ 𝐸𝑥𝑐ℎ𝑎𝑛𝑔𝑒, 𝐺𝐷𝑃,… )
• Notice : That each of these factors are not necessarily drivers (Like
company profit / …) of the share value directly, but merely
correlated with the value of the shares.
• They, in a sense, provide a proxy for potential deviations from
the expected returns of holding the Anglo American share.
• [Note too : Gold Price In this case captures only the deviation
from the expected mean of the price of gold, not its level
changes.]43
Example of APT: Smart Beta
• The advent of Smart Beta products imply investors define certain
characteristics of companies in a systematic way.
• This implies adding additional Betas to the CAPM equation in a structured,
linear way.
• E.g., suppose investors want to invest in the highest Quality (ROE, ROA,
ROIC, etc), best Momentum (say 120day MovingAvg), most stable (lowest
120 SD) best Value (EbitDA / EV) stocks, one could assign scores for each
stock based on these factors, and create a portfolio based on such factors:
• 𝑹𝒌 = 𝑹𝑹𝑭 + 𝜷𝒌 . 𝑹𝑴 − 𝑹𝑹𝑭 + 𝜷𝒌𝑸 + 𝜷𝒌𝑽𝒂𝒍𝒖𝒆 +
𝜷𝒌𝑴𝒐𝒎𝒆𝒏𝒕𝒖𝒎 + 𝜷𝒌 𝑽𝒐𝒍
44
APT
• We will return to APT as an investment vehicle after our
discussions of financial instruments in the next few sessions.
• This will enable us to throw around some of the lingo to get
an understanding of how we can apply some of the insights
provided in the APT framework to designing an actual
investment vehicle:
• not by using immeasurable / abstract factors, but by
specifying actual factor proxies and thereby designing
portfolio weights.
45
Other variants of the CAPM: ICAPM
• The traditional CAPM (when used in a multi-period context) assumes:
• Interest rates and relative commodity prices are fixed (In order for two
period assumption to hold)
• Share returns are i.i.d. Normally distributed
• It isolates consumption risk by the single state variable: the variance in return
• Merton (1973a) extends the CAPM to a continuous time framework
known as the ICAPM.
• It assumes that: Changes in the aggregate pricing risk can be described
by the interest rate → which is assumed the only variable required
to represent changes in the investment opportunity set over time.
46
ICAPM
• Intuitively the ICAPM takes account of the covariance of the asset return
with the market return (as earlier implied by the CAPM, being regarded
as compensation for holding non-diversifiable risk due to changes in the
market)
• BUT also of the asset and the market interest rate (proxied for by the
derivative portfolio [z]).
• The rate of interest can be though of as compensation for the non-
diversifiable changes in consumption risk over time … i.e. changes in
the consumption- or investment opportunity set of consumers that change
over time)
47
CCAPM
• All the variations of the standard portfolio theory done so far
assume that all income is generated from the same source (that of
the investment portfolio).
• CCAPM seeks to remedy this by using instead of the Market
portfolio a proxy for aggregate consumption capabilities of
investors.
• The model assumes that income (in whichever capacity earned) is a
good indication of consumption ability… and therefore the argument
is made to study a portfolio’s returns relative to investors’
consumption abilities (as mimicked by an aggregate income portfolio).
48
Comparison of the variations of MPT pricing
theory models
Model Strengths Weaknesses
CAPM • Single factor identified in the
model
• Simple to use and data easily
available
• Various (mentioned earlier)
ICAPM • Dynamic version of (static)
CAPM, i.e. CAPM through
time
• Nature of factors not specified
in the model
• No other sources of income
CCAPM • Single factor identified in the
model
• Difficult to measure changes in
aggregate consumption
accurately
APT • Not forced to use single factor
• Potentially more accurate
• Nature of factors not specified
in the model.
• This is then especially open to
data-mining problems
49
Alternatives to MPT
• Modern Portfolio Theory and the subsequent asset pricing
models derived from the MVF framework have some serious
shortcomings.
• Yet it is still widely used in practice, despite several other frameworks that
have emerged as more accurate representations of investor behaviour.
• One such framework is the PMPT (Post MPT) framework that
optimizes a portfolio based on returns vs downside risk.
• Risk as defined by variance is a poor proxy for the human
perception of risk – which is the fear of an inferior outcome to
that which is expected (i.e. downside failure).
50
Alternatives to MPT
• Despite the clear benefits to more improved theoretic foundations
possible and better descriptions of risk, the MVF framework is still
the most widely used means of describing investor preferences.
• The question is then why the foundation of finance remains these
simple definitions and are used to guide portfolio decisions.
• Swisher & Kasten (2005) asks this question, and compares it to the
reluctance of the erstwhile medical industry in the 1900s to adopt
the use of sterilization – implying there is always reluctance to
adopt better paradigms, even when we know full well they are
superior...
51
Empirical performance of Asset Pricing Models
• In addition to some of the shortcomings mentioned in the
previous session regarding CAPM’s stringent assumptions –
several other problems emerged in the literature that
questioned the validity of the models.
• A large body of literature emerged that tested the use of these
models in providing quantitative assessments of the fair price
of equity in the market – and showed that at times certain
inconsistencies and anomalies emerge.
• Some of the problematic assumptions that were subsequently
exposed (in addition to those mentioned earlier) include:
53
How robust are the assumptions? What are
the implications of relaxing them?
Assumption Validity? Implications if we relax the
assumption
Investors are risk averse Very robust The problem is not so much that
investors do not display some aversion to
risk, rather that they can at times and for a
portion of their portfolio welcome risk in
an effort to gain higher returns.
Future consumption is
funded solely by future
returns from portfolios of
securities
Not valid CCAPM developed in attempt to
correct for this assumption
Returns are joint normally
distributed
Not valid –
fat tails exist
(Black Swan
events can &
do happen)
Can’t describe returns in terms of mean
and variance only!
Market returns can and do at times
behave very differently to what its
historical record shows.
54
How robust are the assumptions? What are
the implications of relaxing them?
Assumption Validity? Implications?
Homogenous
expectations
Unlikely due to
costly information
No single risk pricing model can exist –
because all agents have unique pricing
models based on their information sets
(but we care more about the subjective
Efficient Frontiers and CML & SML!)
No constraints on
short
sales/borrowing
Not valid for all
markets/securities
CML becomes non-linear above
market portfolio – can only invest in
risky assets beyond this point
Arbitrage may be limited = presence
of abnormal returns which is
inconsistent with model’s pricing
equation
Presence of a risk-
free security
Not perfectly valid
(We listed 3 reasons
earlier…)
Can adapt the model by “creating” a
RF asset through bundling e.g.
Black’s (1972) zero Beta portfolio
creation.
55
How robust are the assumptions? What are
the implications of relaxing them?
Assumption Validity? Implications?
Taxes and
transaction costs
Not valid Can cause non-unique results, but
only if there are different marginal
tax rates and opportunity costs –
usually ignored in practice…
Two time periods Only if interest rates
and relative prices stay
constant and security
returns are IID (Thus
independent of state) –
very unlikely!
Merton’s (1972) Intertemporal
CAPM (ICAPM) developed to
remedy this problem...
Type of asset does
not matter
Not valid:
Various studies found
certain anomalies (e.g.
size, type, value of
assets do in fact
matter!)
Assets perform different to what
the CAPM predicts after
controlling for Beta – extensions to
the models made to account for
these anomalies (e.g. including
firm size / other indicators )
56
Early empirical tests
• Early empirical tests studied the ability of the models (esp. the CAPM) to explain the risk in
security returns in practice
• Black et al. (1972) sorted NYSE shares from 1931 - 1965 into 10 decile portfolios and
estimated the coefficients on the CAPM equation for these portfolios. They then tested the
following simple hypothesis:
• They found that:
• 𝝀𝟎 > 𝟎 and 𝝀𝟏< (𝒊𝑴 – 𝒊) → ∴ shares with low (high) Betas pay higher (lower) returns
than predicted by the CAPM
• It also found that, compared to other more advanced derivatives of the CAPM, the
simple version’s 𝛽 is still the best performing risk measure.57
holds CAPM theIf: ][ and ]0[
: thathypothesis With the
][
00
10
RFm
kkRFTrue
RR
RR
Early empirical tests continued
• Roll’s (1977) critique:
• As mentioned before, the tests of the CAPM are based on an
assumption that the market portfolio is correctly specified ex ante
• All subsequent empirical tests of the CAPM are therefore joint
tests on the: CAPM & M
• BUT specifying the latter is impossible in practice (i.e. using history) –
so no tests are (or ever can be) valid!
58
Further empirical tests
• Fama and French (1993) also showed that a risk model which includes
measures such as firm size & book to market equity ratios, in conjunction
with Beta, provided the best explanation of the cross-sectional variability of
security returns
• Fama and French, e.g., showed that firms that are small and firms with low P/E
ratios tend to out-perform the market on a risk-adjusted basis, concluding that in
many cases outperformance by assets is uncorrelated with their relative
sensitivity to market movements: 𝛽.
• F&F also suggested that Value stocks (Low P/Book or P/E ratio) tend to
outperform Growth stock (High Expected Earnings, thus often having
high P/E’s) irrespective of the Betas.59
Further empirical tests
• These F&F findings have in subsequent years been repeated across various
time-frames and various markets. It seems to clearly suggest that more
than one systematic risk (in addition to the market risk) impacts the price
of assets, in addition to relative market risk.
• Although F&F’s work did have some criticism – notably that size factors do
not explicitly relate to risk (else small firms would merely unite with
other small firms and thus reduce the size premium), while value factors is
based on equal weighting of small and large firms…. –
• Their later work still defiantly show that: P/E, Size, Debt/Equity,
Price/Book add to the explanation of expected stock returns given by
𝛽… Which is kind of stating the obvious…
60
Pricing Puzzles
• Mehra and Prescott (1985) approached the testing problem in a
fundamentally different way:
• They looked at the relationship between the values of the Constant relative
risk aversion measure (CRRA) and the (constant) rate of time preference
that’s implied in the observed patterns of security returns, and whether they
were consistent with those identified in experimental evidence (Thus are
people as risk averse as the models imply?)
• Well.. they weren’t (not by a long way)!! –Two puzzles emerged:
61
Pricing Puzzles
• The equity premium puzzle – investors require a far larger
premium for holding shares, as compared to bonds, than expected in
theory (CRRA needs to be at least 5 times larger than experimental evidence
suggests it should be)
• The risk free rate puzzle – the observed RF rate is much
lower than the SML would predict (for the minimum req’d rate), or the
CCAPM would predict for a realistic CRRA, and even more so when the
required CRRA (to solve the equity risk premium puzzle) is used.
• Hasan and van Biljon (2009) also study these puzzling inconsistencies
from the SA perspective and find that it delivers similar
inconsistencies.62
Summary
• Despite exhibiting at times poor empirical results (although the
CAPM’s performance have shown some of the best results, even
against far more advanced, technical and intricate pricing models),
the standard MPT theoretic framework and the subsequent
CAPM asset pricing model – remains at the core of finance today.
• Used widely and successfully in practice, one of its major benefits
is its simplicity of calculation and ease of interpretation.
• It provides investors and investment intermediaries with a good
quantitative basis from which to evaluate asset prices.
• Its drawbacks and limitations are clear, yet does not invalidate its
use.
Just… use it with care!63